A2a

Agent Interconnect - Discover and invoke other AI Agents. Use when: need cross-platform collaboration, delegate tasks to specialized agents, or build multi-a...

Security Scan

The skill implements a simple WebSocket-based agent discovery/invocation library that matches its description; it asks for no secrets and has no installer, but it has minor engineering issues and requires care when connecting to arbitrary endpoints.

Purpose & Capability

Name/description (agent interconnect, discover/invoke agents) align with the code and SKILL.md. The code provides Registry, Server, and Client for calling agents over WebSockets, which is expected for this purpose. Requiring python3 only is reasonable.

Instruction Scope

SKILL.md only documents how to register actions, call agents, and discover capabilities. It does not ask the agent to read unrelated files or credentials. The runtime behavior (opening WebSocket connections and sending messages) is consistent with the described function.

Install Mechanism

No install spec (instruction-only) — low risk. However, the Python code imports the third-party 'websockets' package inside Client methods but the SKILL.md and manifest do not declare this dependency; runtime will fail unless 'websockets' is available. No external downloads or unusual installers are used.

Credentials

The skill requests no environment variables, credentials, or config paths. This is proportionate to a networked inter-agent library. There are no hidden env accesses in the code or SKILL.md.

Persistence & Privilege

always is false and the skill does not request special privileges or modify other skills. It does initiate outbound network connections (WebSockets), which is normal for inter-agent communication but increases operational exposure and should be controlled by policy.

Assessment

This skill appears to do what it claims: it provides a small WebSocket-based client/server/registry for calling other agents and does not request secrets. Before installing: - Understand that it will open outbound WebSocket connections to endpoints you supply; do not point it at untrusted servers because messages could include sensitive data. - Ensure the Python 'websockets' package is available in the runtime environment (SKILL.md does not declare this dependency). Without it, the client will fail. - Note minor code issues (e.g., a likely typo/bug in Server.__init__) — the implementation is lightweight but not hardened. If you plan to use it in production, review/patch for input validation, errors, use secure wss:// endpoints, and add authentication/authorization between agents. - If you need guarantees about data handling or access controls, require those features (TLS, auth tokens, allowed endpoint list) before using this skill in sensitive contexts.

🔗 A2A - Agent Interconnect / Agent 互联互通

Enable OpenClaw to collaborate with other AI Agents through standard protocol.

When to Use / 使用场景

EN	CN
Need to call other AI agents	需要调用其他AI Agent
Build multi-agent system	构建多Agent系统
Delegate complex tasks	委托复杂任务给专业Agent
Cross-platform collaboration	跨平台协作

Design / 设计原则

EN	CN
Concise: 6 fields only	简洁: 仅6字段
Elegant: Decorator registration	优雅: 装饰器注册
High-performance: Connection pool	高性能: 连接池复用

Message Types / 消息类型

Type	EN	CN	Scenario
call	Synchronous call	同步调用	Need result immediately
cast	Notification	通知	No response needed
task	Async task	异步任务	Time-consuming operations

Core API / 核心API

from a2a import Server, Client, Registry

# Server / 服务端
s = Server('my_agent')
@s.action('echo')
async def echo(p): return p

# Client / 客户端  
c = Client('caller')
r = await c.call('ws://host:8766', 'action', {})

# Registry / 注册中心
r = Registry()
r.reg(Agent('id','name','ws://ep',{'cap':'rw'}))
agents = r.find('cap')

Examples / 示例

# Discover agents that can draw
发现能画图的Agent

# Delegate task
Ask Coze to draw a panda / 让Coze帮我画一只熊猫

Installation / 安装

npx clawhub install a2a-waai

Author / 作者

• WaaiOn